@inproceedings{kim-cho-2025-goodliar,
    title = "{GOODLIAR}: A Reinforcement Learning-Based Deceptive Agent for Disrupting {LLM} Beliefs on Foundational Principles",
    author = "Kim, Soo Kyung  and
      Cho, Hyunsoo",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/display_plenaries/2025.findings-acl.160/",
    pages = "3076--3101",
    ISBN = "979-8-89176-256-5",
    abstract = "Large Language Models (LLMs) often succumb to adversarial prompts, a phenomenon popularly known as ``jailbreaking.'' While jailbreaking primarily targets short-term noncompliance with predefined policies, we argue that a deeper vulnerability lies in altering an LLM{'}s \textit{fundamental axiomatic beliefs}, such as mathematical or philosophical truths. In this work, we introduce GoodLiar, a reinforcement learning (RL)-based framework that generates deceptive contexts to systematically \textit{rewrite} an LLM{'}s core logical or philosophical understandings. By incentivizing an RL agent to produce persuasive and coherent arguments, GoodLiar aims to induce \textit{persistent} belief shifts, rather than merely influencing immediate judgments of factual truthfulness. {\%}rather than one-off policy breaches. Our approach introduces \textit{DA-ILQL}, a novel offline RL method that extends ILQL by integrating on-policy data and language exploration to enhance the language discovery and optimization. Through extensive evaluations on multiple LLMs, we show that deceptive contexts discovered by GoodLiar consistently outperform simple multi-turn prompting methods."
}